# Re: Bidirectional algorithm question (2.0)

From: schererm@us.ibm.com
Date: Wed Aug 18 1999 - 13:01:31 EDT

hi,

i just checked this example with the bidi algorithm version from bidi 3.0 (which
i am implementing right now).
the result is the same, your conclusion is correct.

to my equally euro-centric eyes , i would say that the dollar sign follows the
text and precedes the number in both logical and visual order, and that should
be fine.

however, the bidi algorithm goes only so far. in cases where it fails to read
the user's mind and produces an unexpected result you should try to insert LRM
or RLM characters to get the behavior that you expect.
try AL AL AL (LRM=L) ET EN EN EN.

also have a look at the new tech report version on the web. it is quite improved
and less ambiguous, although it still lacks most of the "why". the committee is
actually working on version 5 of that report, which may not be ready and
published yet.

best regards,

markus

Markus Scherer IBM Cupertino, CA schererm@us.ibm.com

Nick Clarey <nclarey@syd.asc.corp.mot.com> on 99-08-17 18:12:19

To: Unicode List <unicode@unicode.org>
cc:
Subject: Bidirectional algorithm question (2.0)

Howdy all,

I've got a question regarding the processing of European numbers with
Arabic letters.

Here's the situation;

AL AL AL ET EN EN EN

An example of this might be;

AL AL AL \$ 5 0 0

Running through the bidi algorithm, we get the following...

(P0)
AL AL AL ET AN AN AN

(P3)
AL AL AL N AN AN AN

(N3)
AL AL AL R AN AN AN

(I1)
Levels - 1 1 1 1 2 2 2
Text - AL AL AL \$ 5 0 0

(L2)

5 0 0 \$ AL AL AL

Where the "AL" characters are reversed.

Is this correct, or have I erred? We're just split on whether or not this
is correct behaviour, because it appears strange to my eurocentric eyes.

See ya,

Nick

-----
Nick Clarey, Software Engineer
Motorola Australia Software Centre
ph 61-2-9666-0582 fax 61-2-9666-0579

This archive was generated by hypermail 2.1.2 : Tue Jul 10 2001 - 17:20:51 EDT